Ethernet has emerged as the standard of choice for local area networks. With speeds of 10 Mbps, 100 Mbps, 1 Gbps, and 10 Gbps, Ethernet capacity has grown to meet the need for increased network capacities. Consequently, there is considerable interest by operators to offer multipoint network services over public networks. A multipoint network service is a service that allows each of the customer sites to communicate directly and independently with all other customer sites connected to the network via a common interface.
A new network technology that renders multipoint connectivity services has been introduced recently in U.S. patent application Ser. No. 10/265,621 by Casey. This technology is known as “virtual private LAN service” (VPLS). VPLS is a multipoint Layer 2 virtual private network (VPN) technology that allows multiple sites to be connected over a emulated Ethernet broadcast domain that is supported across, for example, multi-protocol label switching (MPLS) networks. That is, VPLS provides connectivity between geographically dispersed customer sites across metropolitan area networks (MANs) or wide area networks (WANs), seemingly as if the customer sites were connected using a LAN.
Abstractly, a VPLS can be defined as a group of virtual switch instances (VSIs) that are interconnected in a full-mesh topology to form an emulated LAN. Specifically, a full-mesh of transport tunnels (i.e pseudowires) needs to be established between network elements (NEs) participating in a single VPLS. Concretely, a VSI can be seen as a bridging function, in which a packet is switched based upon its destination address “DA” (e.g., a medium access layer (MAC) address) and membership in a VPLS. If the destination address is unknown, or is a broadcast or multicast address, the packet is flooded (i.e., replicated and broadcasted) to all transport tunnels associated with the VSI. All NEs participating in a single VPLS instance appear to be on the same LAN.
Reference is now made to FIG. 1, which shows a VPLS 100 established between sites 110, 112, 114 and 116 of a customer A. Sites 110, 112, 114 and 116 are served by NEs 120, 122, 124 and 126 respectively. Each site is connected to an output port of each NE through a customer edge (CE) device (not shown). The VPLS is formed by a full-mesh of pseudowire (PW) connections 130 that interconnect NEs 120 through 126. The PW connections are carried over a MPLS network 150.
NEs in VPLS 100 need to support a “split-horizon” scheme in order to prevent loops. Namely, a NE in VPLS 100 is not allowed to forward traffic from one PW to another PW in the same VPLS. Furthermore, each NE in VPLS 100 needs to implement basic bridging capabilities, such as flooding packets and replicating packets, as well as learning and aging (to remove unused) destination addresses. A packet received at a source NE (e.g. NE 120) is transmitted to its destination based on the DA designated in the packet. If the source NE (in this case 120) does not recognize the destination NE associated with the DA, the packet is flooded to all other NEs in VPLS 100.
A packet to be flooded is replicated in as many copies as the number of PWs 130 connected to a NE, namely, a packet is replicated on all connections that are associated with a particular VSI. The number of packet replications increases linearly as the number of connections in the VSI increases. This replication is not as efficient as the mechanism for transmitting broadcast traffic with Ethernet switching technology, in which broadcast traffic is transmitted only once per physical interface. Furthermore, replicating packets and transmitting them at wire speed may not be feasible.
VPLS and other multipoint network services are designed to operate properly only in a full-mesh logical topology. This significantly limits the bandwidth utilization when providing VPLS over networks that have a network topology other than a full-mesh topology, such as ring topology networks. Therefore, it would be advantageous to allow multipoint network services to be efficiently executed over networks having a topology other than a full-mesh topology. It would be further advantageous to eliminate the shortcomings resulting from the need to replicate packets.